Increasing Efficiency of Routine Robot Space Operations through Adjustable Autonomy and Learning from Human Instructions

Our work on this project has led to contributions along two research thrusts: Robot Manipulation Alongside and in Collaboration with People. Autonomous robot manipulators have the potential to significantly impact a wide range of domains, from households to remote planetary exploration. Across these domains, robots will frequently work among humans, necessitating support for a spectrum of manipulation behaviors ranging from teleoperation to full autonomy. Our work shows that improved autonomy, adjustability of behavior, and adaptiveness to people lead to greater robot efficiency and effectiveness in manipulation tasks when operating alongside and in collaboration with people. Specifically, this research has made the following contributions:
•Human-in-the-loop grasp pose specification for teleoperation: development of novel, latency-robust teleoperation interfaces that offload some of the human’s responsibilities to the robot, and demonstration of the interfaces’ effects on manipulation task performance.
•Pairwise ranking for autonomous grasp calculation: development of an autonomous grasp calculation approach based on pairwise ranking that can adapt to novel objects and be re-trained with few data samples.
•Active perception for autonomous observation: formalization of autonomous human observation as a constraint-based multi-objective optimization problem that, solved in real-time and with the incorporation of robot grasping and human motion prediction, enables unobtrusive recording of humans in isolated environments.
•Adaptive and collaborative task planning: development of a task planning approach using interconnected hierarchical task models that enables real-time collaborative assembly between a human worker and a robot assistant. 
Facilitating Reliable Autonomy with Human-Robot Interaction Autonomous robots are increasingly deployed to complex environments in which we cannot predict all possible failure cases a priori. Robustness to failures can be provided by humans enacting the roles of:
• developers who can iteratively incorporate robustness into the robot system,
• collocated bystanders who can be approached for aid, and
• remote teleoperators who can be contacted for guidance.
However, assisting the robot in any of these roles can place demands on the time or effort of the human. This thesis will develop modules for use during failure interventions leading to increased reliability of autonomous robots while reducing the demand on humans. In pursuit of that goal, this thesis makes the following contributions:
•A development paradigm for autonomous robots that separates task specification from error recovery. The paradigm reduces burden on developers while making the robot robust to failures.
•Development of a model for gauging the interruptibility of collocated humans.
•A human-subjects study shows that using the model can reduce the time expended by the robot during failure recovery
•A human-subjects experiment on the effects of suggestions provided to remote operators during failure interventions. We show that humans need both diagnosis suggestions and action recommendations as decision support during the intervention.